The change we propose is to only require that all nodes are up if cqlmigrate needs to make any changes at all i.e. creating the keyspace, creating the schema migration table and any application specific migrations that have not been applied (determined by querying schema_migrations table)
At the point we determine that a change is needed, the ClusterHealth.check() would be invoked and if this fails the cqlmigrate would abort.
The change we propose is to only require that all nodes are up if cqlmigrate needs to make any changes at all i.e. creating the keyspace, creating the schema migration table and any application specific migrations that have not been applied (determined by querying schema_migrations table)
At the point we determine that a change is needed, the ClusterHealth.check() would be invoked and if this fails the cqlmigrate would abort.
These are the changes agreed in https://github.com/sky-uk/cqlmigrate/pull/36